SYSTEM Cited by 1 source

Vercel AI Gateway¶

Vercel AI Gateway (vercel.com/ai-gateway) is Vercel's model-provider abstraction layer over which AI SDK applications dispatch to arbitrary model providers (OpenAI, Anthropic, Google, Meta, plus others) via one surface.

(Source: sources/2026-04-21-vercel-build-knowledge-agents-without-embeddings)

Why it shows up on this wiki¶

Canonical transport layer for the complexity- tiered model selection pattern as instantiated in Vercel's Knowledge Agent Template. The template's complexity router classifies incoming questions and dispatches each to the appropriate tier; AI Gateway is the abstraction that makes any model slot into either tier without touching the router code.

Verbatim framing from the post:

"The template also includes a smart complexity router. Every incoming question is classified by complexity and routed to the right model. Simple questions go to fast, cheap models. Hard questions go to powerful ones. Cost optimization happens automatically, with no manual rules. Compatible with any AI SDK model provider via Vercel AI Gateway."

The compatibility-with-any-AI-SDK-provider claim is the architectural payoff: the complexity router doesn't hard-code model identities; it routes to tier labels, and AI Gateway resolves tier labels to concrete providers at request time.

Role in the pipeline¶

Every agent model call in the template passes through AI Gateway.
The complexity router picks the tier per question; AI Gateway dispatches to the tier's configured provider.
Provider substitution (e.g. swap the fast/cheap tier from GPT-4o-mini to Claude Haiku) requires zero code change — it's a Gateway-config change.

Adjacent gateway systems on the wiki¶

systems/cloudflare-ai-gateway — Cloudflare's Worker-hosted AI-provider fan-out.
systems/instacart-ai-gateway — Instacart's internal multi-provider gateway.
systems/unity-ai-gateway — Unity's engine-side AI dispatch surface.

All four sit in the same architectural slot (multi-provider LLM fan-out surface) but at different altitudes (hyperscaler / vendor / internal-corp / engine-runtime). Vercel AI Gateway is the vendor-hosted AI-SDK-native instance.

What's undisclosed¶

Gateway overhead / added latency vs direct provider calls.
Failover / retry model across providers.
Observability surface (tracing, logging, cost per request).
Rate-limit / quota policy across provider fan-out.
Authentication model (per-provider credentials vs Gateway-issued token).

Seen in¶

sources/2026-04-21-vercel-build-knowledge-agents-without-embeddings — canonical transport layer for the complexity router; named as "compatible with any AI SDK model provider via Vercel AI Gateway."